Is AI transformation a software purchase?

AI transformation is a decision to rebuild internal information and workflow, and a software license is only one line item in that decision.

Budget requests for AI tools often reach the board in a familiar form: a procurement line item. A software license is added, monthly cost per user is calculated, and security approval is requested from IT. This approach is the same one long used in enterprise software procurement. Yet when moving from the boardroom to operational floors, it becomes clear that AI does not settle into a company like traditional software.

Why is AI not purchased like traditional software?

Traditional enterprise software fixes the process, while AI generates outputs and suggestions within the process.

An ERP or CRM implementation is also a serious organizational effort, but once the system is installed, the process is clear and the steps are fixed. The software follows these steps, and the employee conducts work accordingly. For this reason, buying decisions can largely be made based on price, security standards, and feature lists.

AI models work probabilistically and may not produce the exact same output for the same input every time. They write a draft, present an analysis, or suggest a next step. Therefore, what determines the result is how the business process receiving the tool is structured and how information flows through that process.

What changes in the company when licenses are distributed?

Licenses distributed without redesigning the process often speed up individual tasks while leaving the workflow untouched.

Consider a company with fifty to five hundred employees. Management assigns language model licenses to sixty people and assumes transformation has begun. A few months later, three scenarios often emerge: some licenses are never opened, some are used sporadically for individual tasks like email drafts and meeting summaries, and the rest are added as a new layer on top of the existing process.

In all three scenarios, the foundation of the workflow remains the same. Some steps get faster, but the transfer of information between departments, approval sequences, and responsibility distribution remain unchanged.

Do adoption rates reflect the depth of transformation?

Official adoption figures record usage based on self-reporting by companies, so they do not measure the depth of transformation.

TurkStat's AI Statistics bulletin records whether a company states it uses AI. Usage frequency, how many employees use it, and whether it has passed the pilot phase are not measured separately in this indicator.

The same bulletin highlights another detail. Among enterprises not using AI but considering it, 74.2% cite the lack of relevant expertise within the enterprise as a barrier, while 67.4% point to high costs (TurkStat, AI Statistics 2025, October 1, 2025). Lack of expertise ranks ahead of cost, suggesting that the primary issue lies in human capital and knowledge.

The detailed figures can be found at [Link to article: How many companies in Turkey use AI?].

Where does the bottleneck form in a company?

In mid-sized companies, information often accumulates, alters, and slows down at handoff points between departments.

Consider a sales team forwarding a technical request from a client to the product team to prepare a custom proposal. The sales representative takes meeting notes, enters them into internal systems, and sends an email. The technical specialist reviews the form, asks questions about missing information, and prepares a draft.

When an AI tool is added to this chain, the time the sales representative spends writing emails drops. However, approval steps between departments and the need to verify information remain intact.

Therefore, the question to ask when making an AI decision goes beyond tool selection: "How do we redesign the flow of information and approval steps in our proposal preparation process so that AI can take inputs and generate reliable outputs?"

Who should hold responsibility?

In a process-oriented AI decision, responsibility rests with the leader who directly manages the operation.

A tool-oriented decision shifts responsibility to the software vendor or the IT department. AI projects often stall precisely because of this shift in accountability. The leader thinks they are making a technology purchasing decision, when in fact it is an organizational design decision.

How is the workflow rebuilt?

The workflow is rebuilt by mapping where information originates, where it is transformed, and where it is audited.

Once this map is drawn, certain steps and handoff points prove redundant and are removed from the flow. AI is assigned a clear role as the step generating the initial draft or analyzing data. Employees then transition from producing information from scratch to auditing, correcting, and converting outputs into decisions.

When this role change is not explicitly managed, employees view AI as a threat or an extra task tacked onto their daily work. They maintain their old way of working while taking on the added burden of prompting the tool and correcting its output, increasing their workload instead of reducing it.

At Anthifel, every engagement begins with a Discovery Sprint, clarifying roles, authority, and processes before selecting tools.

When AI is treated as a procurement item, it simply adds a new monthly bill to the company. When the way work gets done is rebuilt, the company's speed of decision-making and output generation can permanently change.

Which three questions should be asked when a budget request arrives?

Before discussing tool features and license fees, three process-related questions should be put to the team making the request.

• Which approval step or interdepartmental handoff in which business process does this tool eliminate?
• When responsibility for producing output shifts to the tool, how will affected employees' time be redirected toward auditing and quality control?
• Where does information currently slow down in the process? Does this tool directly resolve that bottleneck, or does it bypass it and add a new step?

If the answers do not go beyond speeding up the current process, the decision remains merely a tool decision. A process decision begins with the answers to these three questions.